Full text of DOI:10.1207/S15324796ABM2402_09

Audience Status Moderates the Effects of Social Support and Self-Efficacy
on Cardiovascular Reactivity During Public Speaking
Clayton J. Hilmert, M.A., Nicholas Christenfeld, Ph.D., and James A. Kulik, Ph.D.
University of California, San Diego
ABSTRACT
ences, that are associated with unusually large cardiovascular
responses. This work also examines manipulations or interven-
tions that buffer the cardiovascular response. Along these lines,
people have examined the cardiovascular consequences of
high-stress jobs (11,12) and low levels of task control (13–17).
One situational factor that has received significant attention
for its influence on reactivity is the presence of other people.
Kamarck, Manuck, and Jennings (18) showed that the presence
of a friend can reduce the reactivity of a person performing men-
tal arithmetic and concept-formation tasks. This finding sug-
gests possible mechanisms for the robust benefits of social sup-
port uncovered in the epidemiological literature, where the
presence of social support has been reliably linked to decreases
in cardiovascular disease, hypertension, and mortality (19,20).
Therefore, people who are frequently in the presence of friends
and family may show reduced cardiovascular reactivity to
stress, and thus be less likely to suffer later cardiovascular dis-
ease. Since the Kamarck et al. (18) findings, researchers have in-
vestigated the factors that moderate this social support effect.
Not surprisingly, one important factor is the behavior of the per-
son serving as the audience, with positive responses reducing re-
activity more than negative responses (21). Other, less obvious
factors also appear to moderate social support effects. For exam-
ple, support from a woman seems to buffer reactivity, compared
to nonsupport, but identical behavior from a man does not (22).
This finding also fits with the epidemiological evidence sug-
gesting that having a wife provides greater health benefits than
having a husband (23–26).
Recently, we identified another situational factor that ap-
pears to moderate the effects of social support on cardiovascular
reactivity and suggested that an important component may be a
performer’s effort. Supportive behaviors from an audience,
compared to neutral feedback, reduced the reactivity of a person
giving an impromptu speech only when the experimenter was
present during the speech task (27). When the experimenter was
absent, exactly the same supportive behaviors from the audience
increased reactivity. In other words, the usual cardiovascu-
lar-dampening effects of a supportive audience only emerged
when a potentially evaluative experimenter was present. We hy-
pothesize that, in the absence of an evaluative experimenter, the
supportive behaviors served instead to motivate the speaker to
greater effort, and this effort was reflected in greater reactivity.
Without an evaluative experimenter, and with the audience ap-
pearing relatively neutral and disinterested, the speaker may not
have been trying very hard, with relatively low reactivity the
result.
Exaggerated blood pressure responses to stress are impli-
cated in the development of cardiovascular disease, and an ef-
fort has been made to identify factors associated with such re-
sponses. One situational factor that impacts cardiovascular
responses to stress is the presence of other people and their be-
havior. Here, we manipulated the status of the audience during a
stressful public speaking task to explore its impact on reactivity
and its possible role in moderating the effects of the speaker’s
confidence and the audience’s response during the speech.
Sixty-four normotensive female undergraduates, classified as
having high or low self-efficacy for public speaking, gave a
5
-min speech to an audience that responded positively or nega-
tively. Half of the audiences were presented as public speaking
experts and half as novices. Cardiovascular reactivity was
greater for low-efficacy speakers and for those receiving posi-
tive feedback. Reactivity was also greater facing an expert audi-
ence. Furthermore, the effects of both self-efficacy and audience
feedback were intensified before an expert audience. To under-
stand social support effects, we must attend not only to charac-
teristics of the recipient but also to those of the provider.
(
Ann Behav Med
2002, 24(2):122–131)
INTRODUCTION
The reactivity hypothesis suggests that unusually large or
frequent blood pressure responses to stress are associated with
later cardiovascular disease (1,2). The bulk of the research ex-
ploring this notion is concerned with individual differences,
identifying the people who show excessive reactivity and are
thus at elevated risk of later disease. Work in this vein has sug-
gested that the people who show the largest responses to a stan-
dardized stressor are more likely to develop hypertension (3,4)
and are also more likely to have a family history of hypertension
(
5). In addition to linking excessive reactivity to later disease, or
to markers of later disease, the individual-difference approach
has sought to identify factors that are predictive of excessive re-
activity. For example, people high in hostility may show greater
reactivity, especially to anger-relevant provocation (6–8). Simi-
larly, people with low self-efficacy for a given task can show
higher levels of reactivity when performing that task (9,10).
The other major line of work based on the reactivity hy-
pothesis explores the situations, rather than individual differ-
Reprint Address: N. Christenfeld, Ph.D., Department of Psychology
0
109, 9500 Gilman Drive, University of California, San Diego, La
Another factor that may affect how much effort speakers
put into the task, and that may moderate social support effects, is
the status of the audience. In this experiment we explore the role
Jolla, California 92093–0109. E-mail: nicko@ucsd.edu
©
2002 by The Society of Behavioral Medicine.
122

Volume 24, Number 2, 2002
Audience Status and Reactivity
123
that this characteristic of the audience can play on reactivity.
Based on our previous work (27), and other work showing that
an increase in the potential for evaluation can increase reactivity
with low self-efficacy should be especially pronounced before
an audience of experts. In this situation, such heightened reac-
tivity may be due to increased anxiety, increased effort, or some
combination of the two, as the relation between emotion, moti-
vation, and cardiovascular reactivity is anything but simple
(19,30,33,34). Another possible outcome could be that, for
someone with low confidence in speaking ability, any audience
is frightening, but confident speakers only get nervous in front
of experts. In this case, we might find the effects of self-efficacy
on reactivity would only be visible in front of a novice audience.
And, again, it is possible that any effects of self-efficacy and this
particular characteristic of the situation would be independent,
with audience expertise playing no moderating role.
Overall, this work aims to extend our knowledge of the fac-
tors that influence the cardiovascular response of a person under
stress, by exploring a new situational factor that might also serve
to moderate other documented influences on reactivity. Spe-
cifically, the effects of both social feedback and self-efficacy on
the reactivity of a speaker may be greater when public speaking
experts are the audience for that performance. To explore these
possibilities, we examined the cardiovascular response of peo-
ple giving an impromptu speech to a 2-person audience. Half of
the speakers were categorized, based on self-report assessment,
as having high self-efficacy for public speaking and half as hav-
ing low self-efficacy. For half of the speakers, the audience be-
haved in an encouraging, supportive manner, and for half, the
audience was more disapproving and disinterested. Finally, half
of the speakers were lead to believe that the 2 people listening to
the speech were expert and experienced public speakers, and
half believed that the audience had no particular public speaking
skill or experience. In all cases, the experimenter remained out
of the room, and out of earshot of the speech itself.
(
28–30), we expect that an audience composed of public speak-
ing experts might engender greater reactivity in a speaker than
would an audience of novices. In addition to this effect, it is pos-
sible that the expertise of the audience could interact with, or
moderate other factors that influence reactivity during a stress-
ful task. Two other possible factors are considered: (a) a situa-
tional manipulation—the supportiveness of the audience and (b)
an individual difference—the self-efficacy of the speaker.
In this experiment, we again focus on social support in the
absence of the experimenter, in an effort both to replicate the
previously reported “backwards” social support effect (where,
with the experimenter absent, support is associated with in-
creased reactivity) and to examine its possible interaction with
audience expertise. One possibility is that an audience of public
speaking experts is simply more important than one made up of
those with no public speaking experience, and so any effect of
an audience would be magnified. In this case, we should find
that the increase in reactivity that accompanies support should
be greater with an expert audience. In other words, people
should show more effort when they get positive feedback and
care more about the feedback of experts; they should show an
especially large increase when the experts seem to be approving.
Another possibility is that, although disinterest from novices
could cause speakers to reduce their effort, similar feedback
from experts would be highly motivating. Impressing experts
may be more important, and so evidence that one is not yet suc-
cessful could spur renewed effort and, thus, heighten reactivity.
In this case, one might find that with a novice audience, positive
feedback produces high reactivity, but with experts, it is the neg-
ative feedback that engenders greater reactivity. Another possi-
bility, of course, is that the behavior of the audience and the sta-
tus of that audience will exert independent effects on a speaker’s
cardiovascular response.
We also examine whether a characteristic of the audience,
in this case expertise, interacts with an individual-difference di-
mension, in this case self-efficacy. Several studies have shown
that people’s reactivity while performing stressful tasks is influ-
enced by their subjective impressions of their own competence
at that task (9,10,31,32). In a study by Gerin, Litt, Deich, and
Pickering (31), for example, they found that low self-efficacy
for a mathematics task was associated with greater reactivity,
but this effect was only apparent when participants could control
the timing of the mathematics problems. In other words, the po-
tential cost of being a person with low self-efficacy, in terms of
suffering greater cardiovascular reactivity, may be amplified in
some situations, in this case, by having greater control over out-
comes.
METHOD
Participants
Sixty-four undergraduate women students attending the
University of California, San Diego, participated for class
credit. All participants were normotensive (resting blood pres-
sure less than 140/90).
Cardiovascular Recording
Blood pressure and heart rate were continuously assessed
using an Ohmeda Finapres 2300 Blood Pressure Monitor (Oh-
meda, Louisville, CO). This apparatus takes beat-to-beat mea-
sures of systolic blood pressure (SBP), diastolic blood pressure
(DBP), and heart rate (HR) noninvasively, using a finger cuff
worn on the third finger of the nondominant hand. Several au-
thors have described the method of operation of the Finapres
and its reliability, especially for tracking rapid cardiovascular
changes (35–37).
One situational factor that could amplify the effects of hav-
ing low efficacy for public speaking is facing an audience of
public speaking experts. Just as we can test whether the exper-
tise of the audience amplifies the effects of that audience’s feed-
back, we can also explore whether it amplifies the effects of a
speaker’s confidence. In this case, the greater reactivity of those
Procedure
Participants were told that the study concerned blood pres-
sure and heart rate changes while performing a task. Once the
operation of the Finapres was described, the participant signed a
consent form and she was fitted with a finger cuff. The partici-

1
24
Hilmert et al.
Annals of Behavioral Medicine
pant then was informed she would perform a short speech. She
was told that the topic of the speech would be assigned, that she
would have a 5-min preparation period, and that the speech
would be given into a microphone and to an audience of two stu-
dent volunteers recruited by the experimenter’s advisor.
The participant then spent 5 min preparing her speech. She
was supplied with a pen and paper and was instructed that her
speech should agree or disagree with the statement “College is a
valuable asset” and that she should outline some ideas concern-
ing what she wanted to say. The experimenter left the room for
the preparation period.
After 5 min, the experimenter returned and asked the partic-
ipant to complete a self-efficacy questionnaire (described later
in this article). Once this was completed, the experimenter casu-
ally described the two people who would be serving as the audi-
ence. They were described either as public speaking experts or
novices, depending on condition (described later in this article).
The participant was then asked to sit as still as possible while
baseline measures of HR and blood pressure were taken, and the
experimenter left the room for 3 min. This is shorter than tradi-
tional resting baselines. However, because participants had
plenty of time to become accustomed to the laboratory before
the baseline was assessed, and because of the very high number
of readings provided by the Finpres, 3 min should be sufficient
to provide a highly stable measure (37).
rienced in public speaking, that both had been on speech teams
in high school, that one of them was currently coaching debate
at a local high school, and that “We are very lucky to have them
help us today.” In novice-audience conditions, the participant
was told that the audience had very little experience in public
speaking and was just made up of two undergraduate psychol-
ogy majors who were nice enough to volunteer their time. The
audience members themselves remained blind to the expert or
novice condition.
Feedback. Participants were randomly placed into one of
two feedback conditions. In the positive-feedback condition,
both confederates remained neutral but attentive for the first 45
sec and then started to indicate interest and approval, with one
confederate playing a more active role. Both confederates
leaned forward with an open body posture, smiling and nodding
in agreement with the participant’s speech. The more active au-
dience member occasionally gave verbal approval of the speech
by making comments such as “good point,” or “that’s right,” and
the other audience member would agree with such comments.
Approximately 2 min into the speech, the confederates looked at
each other and smiled in agreement that the participant’s speech
was impressive. For the remaining 3 min of the speech, the con-
federates continued to nod, smile, and express verbal agreement
with points made by the speaker.
After the experimenter returned, two women confederates
knocked on the laboratory door and were greeted by the exper-
imenter. Once the audience was seated, the experimenter told
them that they were just to listen to the speech as if they were
listening to anyone giving a talk. Then the participant was re-
minded that her speech had to last 5 min and that it was impor-
tant that she keep her nondominant arm, with the finger cuff,
as still as possible. After the participant said she understood,
the experimenter began the audiotape recorder, unobtrusively
started a timer that was visible to the audience but not the par-
ticipant, and told the participant to begin. The confederates
used the timer to coordinate responses to the speech (as de-
scribed later in this article). The experimenter then left the
room for 5 min.
In the negative-feedback condition, confederates began re-
sponding 45 sec into the speech by leaning back in their chairs
and folding their arms. The more active confederate was atten-
tive but appeared to question the speaker’s conclusions by fur-
rowing her brow occasionally, whereas the other confederate
glanced around the room and showed little interest in the speech.
Two minutes into the speech, the confederates turned to each
other and conveyed, by shaking their heads slightly and sighing,
that neither was very interested in the speech. During the re-
maining 3 min, the confederates showed very little facial expres-
sion and glanced periodically around the room. Additionally, at
3
min one confederate looked at her watch, and at 4 min the
other stifled a yawn. The experimenter remained blind to feed-
back condition throughout the experiment.
After 5 min, the experimenter returned, excused the audi-
ence members, and removed the blood pressure cuff. The partic-
ipant then completed the posttask questionnaire, which con-
sisted of 12 Likert-type questions and included manipulation
checks and ratings of stress during the speech task (described
later in this article).
Measures
Self-efficacy. The self-efficacy questionnaire asked the
participant a total of 16 questions about her confidence in public
speaking. Half of the questions asked about her confidence that
she could deliver a speech calmly, and half asked whether she
could deliver one that was well organized. Additionally, half
asked about speeches on a topic of her own choosing, and half
asked specifically about her confidence giving a speech on
whether college was a valuable asset (the topic of the upcoming
talk). Finally, within each sort of question, she was asked about
her confidence delivering the speech to four different sorts of
audiences: a friend, a stranger, 5 to 10 people, and 25 to 50 peo-
ple. There were thus 16 permutations of these questions, with
each one rated on a 10-point scale from 1 (not at all confident) to
10 (very confident). For example, participants were asked to rate
how confident they were that they could calmly deliver a speech
After the posttask questionnaire was completed, the partici-
pant was debriefed and thanked for her participation.
Experimental Manipulations
Audience expertise. The expertise manipulation was im-
plemented while the participant and experimenter were waiting
for the two audience members to arrive. The experimenter, ap-
parently just filling time, casually mentioned his advisor’s de-
scription of the two people who would be listening to the
speech. In expert-audience conditions, the experimenter said
that his advisor had mentioned that the audience was very expe-

Volume 24, Number 2, 2002
Audience Status and Reactivity
125
about whether college was a valuable asset to an audience of 5 to
Preliminary analyses involved a 2 (self-efficacy classifica-
tion) × 2 (audience expertise) × 2 (feedback) analysis of vari-
ance (ANOVA) performed on baseline measures and Pearson
correlations computed between change scores and baseline
measures. If there were a significant relation between baseline
and condition, or baseline and reactivity, then baseline would be
used as a covariate in the primary analyses. When there was a
significant relation, this procedure would statistically control
for variations in reactivity due to baseline rather than the speech
task (40).
The primary analyses used a 2 (self-efficacy classification)
× 2 (audience expertise) × 2 (feedback) multivariate analysis of
covariance (MANCOVA) to examine the effects of experimen-
tal conditions on SBP, DBP, and HR reactivity simultaneously.
Because previous literature has suggested that simply talking
can raise blood pressure (41,42), word production was incorpo-
rated into the analyses as a covariate. If the multivariate test sug-
gested a significant effect, then paired comparisons were com-
puted using one-way analysis of covariance (ANCOVA)
procedures with word production as a covariate.
1
0 people. This format is similar to those used by Bandura (38)
and Gerin et al. (31) to assess self-efficacy. High reliability was
found for the self-efficacy questionnaire items (α = .96). The
self-efficacy ratings were therefore summed to create a self-effi-
cacy index, with higher values indicating greater self-efficacy.
Scores ranged from 32 to 142 with a standard deviation of 28.
Participants were classified as having high self-efficacy if they
fell above the median of 88 and low if they fell below it. The 2
participants who had self-efficacy scores of 88 were randomly
assigned a low- or high-self-efficacy classification (1 low and 1
high).
Self-reported anxiety. Four items on the posttask question-
naire asked the participant to rate how pleasant and how stress-
ful the speech task had been and how nervous and how calm she
had felt on a scale from 1 (not at all) to 5 (very much). The four
items were reverse scored when appropriate and summed to cre-
ate an anxiety index (α = .90). Higher values on the self-reported
anxiety index indicate greater anxiety.
Self-reported measures, including manipulation checks and
ratings of anxiety, were analyzed using 2 (self-efficacy classifi-
cation) × 2 (audience expertise) × 2 (feedback) ANOVAs. Also,
because word production can provide a measure of how much
effort a participant put into the task, it was examined with a 2 × 2
Manipulation checks. Participants rated their impressions
of the audience’s expertise in public speaking on a 5-point scale
ranging from 1 (not at all) to 5 (very much). However, comments
during debriefing led us to suspect there were interpretation
problems with this item. Therefore, for the final 16 participants,
three questions were added to the posttask questionnaire, which
asked specifically for ratings of the audience’s previous knowl-
edge, training, and experience in public speaking (using the
same 5-point scales). A total “perceived audience expertise” in-
dex was computed by summing these items (α = .91). Higher
values indicate that the participants perceived the audience as
having more public speaking expertise.
To check the effectiveness of the audience-feedback manip-
ulation, six posttask questions asked participants to rate their
impressions of how well the audience attended to the speech,
how emotionally supportive and friendly the audience was, how
much the audience approved of the performance and content of
the speech, and how nervous the audience’s responses made the
participant. Once again 5-point scales were used ranging from 1
×
2 ANOVA. In addition, mediation analyses were conducted to
explore the possible role of effort, as indexed by word produc-
tion, in mediating the effect of audience feedback. This analysis
allows us to see if word production (i.e., effort) is able to account
for the relation between the independent variable and cardiovas-
cular reactivity (43).
RESULTS
Manipulation Checks
Audience expertise index. A 2 (self-efficacy classification)
× 2 (audience expertise) × 2 (feedback) ANOVA performed on
the initial audience-expertise manipulation check item con-
firmed our suspicions concerning the misinterpretation of the
posttask question. We found only a main effect of feedback, F(1,
56) = 8.16, p < .01. Participants reported feeling more positively
about the audience’s public speaking knowledge when the audi-
ence was positive (M = 3.0) compared to negative (M = 2.5). Al-
though this result indicates that participants were aware of the
feedback manipulation, it does not tell us how their perception
of audience expertise was affected.
(
not at all) to 5 (very much). A total “perceived feedback” index
was then computed by summing the 6 feedback-manipulation
check answers (α = .90). Higher values on this index indicate
that participants felt that the audience offered more positive
feedback.
Examination of the supplementary perceived expertise in-
dex (see Method section) did give us some indication that the ex-
pertise manipulation was effective. Because of the small number
of participants who received the supplementary items (n = 16), a
t test was used. This revealed a marginally significant result,
with audiences rated as having more expertise in public speak-
ing in the expert conditions (M = 9.5) than in the novice con-
ditions (M = 7.0), t(14) = 1.96, p = .07. Separate t tests revealed
no other significant effects (ps > .08). Thus, the audience exper-
tise manipulation appears to have been effective. Further evi-
Word production. Audiotapes of the speeches were tran-
scribed, and the number of words uttered during each speech
was counted.
Statistical Analyses
Cardiovascular reactivity was assessed by subtracting aver-
age levels of SBP, DBP, and HR during resting baseline from
average levels during the participant’s speech. Means were cal-
culated using the pulse-based average (39).

1
26
Hilmert et al.
Annals of Behavioral Medicine
dence of this was found when we examined the perceived feed-
back index.
Perceived feedback index. A 2 (self-efficacy classifica-
tion) × 2 (audience expertise) × 2 (feedback) ANOVA was per-
formed on the perceived feedback index. The results indicated a
highly significant feedback condition effect, with higher (more
positive) ratings in positive-feedback conditions (M = 24) rela-
tive to negative conditions (M = 15.3), F(1, 56) = 78.81, p <
.001. The feedback manipulation appears to have been effective.
The ANOVA also showed a significant difference between
expertiseconditions. Theperceivedfeedbackindexwassensitive
enoughtoindicatethat,overall,anexpertaudiencewasperceived
as less positive than a novice audience (Ms = 18.4 vs. 20.8), F(1,
5
6) = 4.55, p < .05. Because the confederates were blind to exper-
tise condition, this cannot reflect actual differences in audience
behaviorbutdoesprovideadditionalevidencethattheparticipant
attended to the audience expertise manipulation.
Preliminary Analyses
Separate 2 (self-efficacy classification) × 2 (audience ex-
pertise) × 2 (feedback) ANOVAs were performed on mean SBP,
DBP, and HR baseline measures. Although participants were al-
ready aware of the nature of the task and the expertise of their
audience, and a difference in anticipatory reactivity could have
been expected, no condition differences were found (all ps >
.10). Correlations also were computed between baseline physio-
logical measures and reactivity measures. Because there were
no significant correlations between baseline and reactivity mea-
sures (ps > .30), and because a lack of association between base-
line and reactivity is common in studies involving normotensive
participants (44), none of the primary analyses used baseline as
a covariate. Reactivity in the primary analyses was determined
by subtracting baseline means from corresponding task means.
Primary Analyses
A 2 (self-efficacy classification) × 2 (audience expertise) ×
(feedback) MANCOVA was performed on SBP, DBP, and HR
reactivity, using word production as the covariate. The results
indicated first that word production showed significant positive
associations with SBP reactivity and DBP reactivity, Fs(1, 55) =
FIGURE 1 Main effects of a) self-efficacy, b) audience expertise,
and c) feedback on systolic blood pressure reactivity (mmHg), diastolic
blood pressure reactivity (mmHg), and heart rate reactivity (BPM).
Error bars indicate standard error of the mean with the covariate
incorporated.
2
1
3.63 and 16.30, respectively, ps < .001 but not with HR reactiv-
ity, F(1, 55) = 1.85, p > .05.
lieved to be experts showed a greater increase in SBP than did
those speaking to an audience believed to be novices, F(1, 55) =
Consistent with previous reports on self-efficacy (9,10),
Figure 1a shows that people who were low in self-efficacy
showed higher cardiovascular responses overall during the task
than those with high self-efficacy, F(3, 53) = 4.92, p < .005. Sep-
arate analyses indicated that the low self-efficacy group was
higher than the high self-efficacy group for SBP reactivity, F(1,
5.49, p < .05. The increase in DBP while talking to experts like-
wise exceeded that produced by talking to novices, F(1, 55) =
5.72, p < .05, as did the increase in HR, F(1, 55) = 12.45, p <
.001.
5
<
5) = 10.80, p < .005, and for DBP reactivity, F(1, 55) = 13.88, p
.001. The means for HR reactivity were similar in direction
Consistent with our previous finding on the effects of social
support when the experimenter is not present during the speech
(27), it can be seen in Figure 1c there was an overall tendency for
greater reactivity among speakers who received positive relative
to negative feedback, F(3, 53) = 2.49, p = .07. In separate analy-
ses, the effect was significant for SBP, F(1, 55) = 7.38, p < .01.
For DBP, the same pattern was present, though it did not quite
but did not differ significantly, F(1, 55) = 1.15, p > .05.
Figure 1b shows that the expertise of the audience also had
an effect on the speaker’s reactivity that was significant across
all three measures, F(3, 53) = 5.88, p < .005. Separate analyses
indicated that participants talking to an audience that they be-

Volume 24, Number 2, 2002
Audience Status and Reactivity
127
reach significance, F(1, 55) = 2.43, p = .12. HR reactivity was in
the same direction, though it did not approach significance, F(1,
novice audience, the difference between positive and negative
feedback was very slight. Individual analyses indicated this in-
teractive effect was marginally significant for SBP, F(1, 55) =
3.67, p = .06, significant for DBP, F(1, 55) = 5.61, p < .05, and
nonsignificant for HR, F(1, 55) = 2.14, p > .10. Simple effects
analyzed with Fisher’s least significant difference method re-
vealed that in expert audience conditions, positive feedback
caused greater SBP and DBP reactivity than negative feedback
(ps < .01). In front of a novice audience, the effect of positive
feedback on reactivity was no different from that of negative
feedback (ps > .10). Because there was not a significant Exper-
tise × Feedback effect on HR reactivity, simple effects were not
analyzed. The other effects from the MANCOVA, the Feedback
× Self-Efficacy interaction, and the three-way interaction did
not approach significance, F(3, 53) = 1.40, p > .25, and F(3, 53)
= 1.23, p > .30, respectively.
5
5) = .35, p > .20.
In addition to expecting the foregoing three main effects,
we also expected that audience expertise might separately inter-
act with the feedback manipulation and the self-efficacy clas-
sification. We found support for both hypotheses. First, the
MANCOVA analysis indicated a significant overall Audience
Expertise × Self-Efficacy effect on reactivity, F(3, 53) = 2.89, p
<
.05. As can be seen in Figure 2, across each measure, when
participants were facing an expert audience, those with low
self-efficacy for public speaking had greater reactivity than par-
ticipants with high self-efficacy, whereas self-efficacy made lit-
tle difference when performing in front of a novice audience.
Separate analyses indicated that this interaction approached sig-
nificance for SBP, F(1, 55) = 3.53, p = .06 and DBP reactivity,
F(1, 55) = 3.08, p = .08, and reached significance for HR reac-
tivity, F(1, 55) = 5.90, p < .02. Of more interest, simple effects
analyzed with Fisher’s least significant difference method (45)
indicated that for each reactivity measure, reactivity was greater
for low self-efficacy than high self-efficacy performers in front
of an expert audience (all ps < .05); however, reactivity of high-
and low-self-efficacy individuals did not differ in front of a nov-
ice audience (all ps > .10).
The MANCOVA results also indicated an overall tendency
for the effect of feedback to vary as a function of audience ex-
pertise, F(3, 53) = 2.51, p = .07. As can be seen in Figure 3,
when participants performed in front of an expert audience, pos-
itive feedback appeared generally to lead to greater reactivity
than negative feedback, whereas when performing before a
Self-Reported Anxiety
A 2 (self-efficacy classification) × 2 (audience expertise) ×
2
(feedback) ANOVA was performed on the self-reported anxi-
ety index. This indicated that low-self-efficacy participants re-
ported experiencing more anxiety (M = 14.2) than did high-
self-efficacy participants (M = 12.2), F(1, 56) = 5.80, p < .05. No
other effect was significant (ps > .25).
Word Production
A 2 (self-efficacy classification) × 2 (audience expertise) ×
2 (feedback) ANOVA was performed on the word production
measure. The only significant effect was that people uttered
FIGURE 2 Mean systolic blood pressure reactivity (mmHg), diastolic blood pressure reactivity (mmHg), and heart rate reactivity (BPM) of high-
and low-self-efficacy participants across audience status conditions. Error bars indicate standard error of the mean with the covariate incorporated.

1
28
Hilmert et al.
Annals of Behavioral Medicine
FIGURE 3 Mean systolic blood pressure reactivity (mmHg), diastolic blood pressure reactivity (mmHg), and heart rate reactivity (BPM) across
feedback and audience status conditions. Error bars indicate standard error of the mean with the covariate incorporated.
more words when they received positive (M = 802.2) as opposed
to negative (M = 710.7) feedback, F(1, 56) = 6.13, p < .05.
Consistent with our previous suggestion that when the ex-
perimenter is absent, social support increases reactivity by in-
creasing effort (27), we found evidence that word production
mediated the relation between feedback and reactivity. A vari-
able is a mediator to the extent that it accounts for the relation
between an independent and dependent variable. Baron and
Kenny (43) suggest three conditions, tested with regressions,
that must be satisfied: the independent variable (here, feedback)
must significantly predict the mediator (here, word production);
the independent variable must significantly predict the depend-
ent variable (here, reactivity); and, when the independent vari-
able and the mediator are entered in a multiple regression, the
mediator must still predict the dependent variable and reduce
the previously significant relation between the dependent and
independent variable. In our regression analyses we found that
for SBP reactivity, with feedback marginally predicting DBP re-
activity when entered alone (β coefficient = .22, SEb = 1.80, p
=.09) and having a considerably reduced effect (β coefficient =
.08, SEb = 1.86, p > .50) when entered with word production,
which did predict DBP reactivity (β coefficient = .35, SEb = .01,
p < .01). These results support our previous suggestion (27) that
with the experimenter absent, positive feedback increases reac-
tivity, at least in part, by motivating participants to put forth
more effort (i.e., produce more words).
DISCUSSION
The experiment replicates some previously reported effects
and also suggests new variables that can influence cardiovascu-
lar responses to stress. As with previous reports (9,10) we found
that people with high self-efficacy for a task can show smaller
cardiovascular responses during that task than people with low
efficacy. People who do not expect to succeed may find their
task more threatening. We also replicated our previous finding
that, with the experimenter out of the room during the speech,
positive feedback from the audience can produce greater physi-
ological responses than negative feedback. We also found evi-
dence consistent with our suggestion that this counterintuitive
finding, the opposite of the more common social support effect,
may be due to positive feedback encouraging greater effort, or
active coping on the part of the speaker, when evaluative
concerns are otherwise relatively low (27). Mediational analy-
ses revealed that the effect of positive feedback on reactivity in
this experiment was partly accounted for by the number of
words participants uttered. That is, positive feedback increased
(
=
a) feedback predicted word production (β coefficient = .39, SEb
34.44, p < .01); (b) feedback predicted SBP reactivity (β coef-
ficient = .34, SEb = 2.66, p < .01); and (c) when SBP reactivity
was regressed on both word production and feedback, word pro-
duction was significantly related to reactivity (β coefficient =
.34, SEb = .01, p < .01) and the link between feedback and SBP
reactivity was reduced (β coefficient = .21, SEb = 2.75, p = .10).
Additional mediational analyses on HR and DBP reactivity
revealed that word production and feedback were not signifi-
cantly related to HR reactivity (ps > .40), which is not surprising
given the ANCOVA results reported previously. However, anal-
yses of DBP reactivity indicated a pattern similar to that found

Volume 24, Number 2, 2002
Audience Status and Reactivity
129
effort, indicated by word production, which in turn increased
reactivity.
gested both by the cardiovascular response and the measure of
speech effort (number of words produced). It may be, though it
is just speculative, that, in front of a novice audience, the reas-
suring and effort-encouraging effects more nearly cancel, but in
front of experts the encouragement dominates. In addition, the
fact that word production was not related to audience expertise
or self-efficacy effects in this experiment suggests that these fac-
tors influence reactivity through something other than an effort
mechanism alone.
We have shown that altering the expertise of other people
present during a stressful experience can have both direct effects
and also moderate other effects on reactivity. It could well be
that the intensification due to high-status observers could apply
to other factors, both individual difference and situational, as
well. For example, it could be that the difference between people
high and low in hostility (6–8) is exacerbated when people are
provoked by, or in front of, a high-status audience. It would also
be worth examining whether, in circumstances where positive
regard reduces reactivity, increasing the status of the person pro-
viding feedback would increase the benefit. That is, if approval
from a low-status person reduced reactivity, would the same ap-
proval from someone of high status reduce it even more?
Increasing the perceived expertise of the audience is just
one way of increasing its status, and others are worth exploring.
People who control real rewards, people of high social position,
and attractive people might produce greater effects. This idea
has been suggested as a possible explanation for the finding that
positive feedback from a friend produces greater benefits than
the same feedback from a stranger (47). In general, the effect of
other people may be intensified when, for whatever reason, we
care especially what those people think of us.
In addition to replicating feedback and self-efficacy effects,
the experiment produced novel findings related to the status of
the audience. Three fairly consistent effects emerged for this
factor. First, overall, when people thought that they were talking
to public speaking experts they showed higher cardiovascular
responses than when they thought they were talking to people
with no such expertise. More interestingly, the expertise of the
audience also moderated the effect of self-efficacy. Thus, the ef-
fects of self-efficacy were exaggerated in front of an expert audi-
ence, relative to in front of novices. Third, the expertise of the
audience moderated the effects of feedback from that audience.
In front of experts, the difference between positive and negative
feedback was greater than in front of novices.
Across all three effects, increasing the status of the audi-
ence intensified each effect on cardiovascular reactivity. This in-
tensification operates both on an individual-difference variable
—
self-efficacy—and on a situational manipulation—the nature
of the feedback. It may be that the speakers think that public
speaking experts are going to be more evaluative than novices,
and they therefore care more deeply about their performance
when experts are watching. Experts have the means to be eval-
uative, and almost by definition, experts are better able to tell
what is right or wrong, good or bad, in a situation. This may be
particularly true when the performer is a novice and therefore
relatively unable to make self-evaluations based on previous
experience (46).
When the audience is presented as having no expertise in
public speaking, it has comparatively little impact. It does not
matter, at least in terms of the cardiovascular response, whether
the speaker has high or low confidence in her speaking ability,
perhaps because expecting to succeed or fail is not important in
front of people who will not know the difference. It also appears
not to matter very much whether an audience of novices ap-
proves or disapproves of the speech, perhaps because it is not as
encouraging to see nods and smiles from people whose opinions
are less important or less diagnostic.
Although such conjectures remain to be investigated, it is
clear that in understanding social support effects, we must attend
tocharacteristicsnotonlyoftherecipientbutalsooftheprovider.
Understanding the circumstances that exaggerate cardiovascular
responsesmay, ultimately, helpwithourunderstandingofthesit-
uations, individual attributes, and interactions between the two
that put people at risk of later cardiovascular disease.
The self-report measures of stress during the speaking task
are not very illuminating, as is often the case with this sort of ex-
periment (19). The only effect that emerged was for self-effi-
cacy. This is not particularly surprising, because the classifica-
tion was made on the basis of their self-reported confidence
about public speaking, and the observed correlation was with
self-reports of anxiety during a public speaking task. It is harder
to explain why the other effects should not be present in the anx-
iety index. The effect of audience expertise, for example, was
readily apparent in blood pressure and HR reactivity, and one
might have expected that people would also be able to report
more anxiety speaking to experts. It may be the case that people
are not willing or able to report reliably on their internal states
REFERENCES
(
(
1) Krantz DS, Manuck SB: Acute psychophysiologic reactivity
and risk of cardiovascular disease: A review and methodologic
critique. Psychological Bulletin. 1984, 96:435–464.
2) Pickering TG, Gerin W: Cardiovascular reactivity in the labora-
tory and the role of behavioral factors in hypertension: A criti-
cal review. Annals of Behavioral Medicine. 1990, 12:3–16.
(3) Light KC, Sherwood A, Turner JR: High cardiovascular reac-
tivity to stress: A predictor of later hypertension development.
In Turner JR, Sherwood A, Light KC (eds), Individual Differ-
ences in Cardiovascular Response to Stress. New York: Ple-
num, 1992, 281–293.
4) Manuck SB, Kasprowicz AL, Muldoon MF: Behaviorally-
evoked cardiovascular reactivity and hypertension: Conceptual
issues and potential associations. Annals of Behavioral Medi-
cine. 1990, 12:17–29.
(
(
19). It is also likely that the blood pressure response is not sim-
ply a marker of anxiety. For example, we believe that in this ex-
periment, blood pressure is responding both to effort and to anx-
iety. Positive feedback may have some reassuring effect, but its
dominant effect seems to have been to increase effort, as sug-
(
5) Fredrikson M, Matthews KA: Cardiovascular responses to be-
havioral stress and hypertension: A meta-analytic review. An-
nals of Behavioral Medicine. 1990, 12:30–39.

1
30
Hilmert et al.
Annals of Behavioral Medicine
(
(
(
6) Davis MC, Matthews KA, McGrath CE: Hostile attitudes pre-
dict elevated vascular resistance during interpersonal stress in
men and women. Psychosomatic Medicine. 2000, 62:17–25.
7) Fredrickson BL, Maynard KE, Helms MJ, et al.: Hostility pre-
dicts magnitude and duration of blood pressure response to an-
ger. Journal of Behavioral Medicine. 2000, 23:229–243.
8) Suarez ECE, Harralson TL: Hostility-related differences in the
associations between stress-induced physiological reactivity
and lipid concentrations in young healthy women. Interna-
tional Journal of Behavioral Medicine. 1999, 6:190–203.
9) Bandura A: Self-efficacy mechanism in human agency. Ameri-
can Psychologist. 1982, 37:122–147.
seh Community Health Study. Journal of Health & Social
Behavior. 1986, 27:62–77.
(26) WiklundI, OdenA, SanneH, etal.:Prognosticimportanceofso-
matic and psychosocial variables after a first myocardial infarc-
tion. American Journal of Epidemiology. 1988, 128:786–795.
(27) Hilmert CJ, Kulik JA, Christenfeld N: The varied impact of so-
cial support on cardiovascular reactivity. Basic and Applied So-
cial Psychology (in press, 2001).
(28) Henchy T, Glass DC: Evaluation apprehension and the social
facilitation of dominant and subordinate responses. Journal of
Personality and Social Psychology. 1968, 4:446–454.
(29) Kors D, Linden W, Gerin W: Evaluation interferes with social
support: Effects on cardiovascular stress reactivity in women.
Journal of Social and Clinical Psychology. 1997, 16:1–23.
(30) Wright RA, Tunstall AM, Williams BJ, Goodwin JS, Har-
mon-Jones E: Social evaluation and cardiovascular response:
An active coping approach. Journal of Personality and Social
Psychology. 1995, 69:530–543.
(
(
10) Bandura A, Reese LB, Adams NE: Microanalysis of action and
fear arousal as a function of differential levels of perceived
self-efficacy. Journal of Personality and Social Psychology.
1
982, 43:5–21.
(
11) Laflamme N, Brisson C, Moisan J, et al.: Job strain and ambula-
tory blood pressure among female white-collar workers. Scan-
dinavian Journal of Work, Environment & Health. 1998,
(31) Gerin W, Litt MD, Deich J, Pickering TG: Self-efficacy as a
moderator of perceived control effects on cardiovascular reac-
tivity: Is enhanced control always beneficial? Psychosomatic
Medicine. 1995, 57:390–397.
24:334–343.
(
(
12) Steptoe A: Psychosocial factors in the development of hyper-
tension. Annals of Medicine. 2000, 32:371–375.
13) Gerin W, Pieper C, Marchese L, Pickering TG: The multi-di-
mensional nature of active coping: Differential effects of effort
and enhanced control on cardiovascular reactivity. Psychoso-
matic Medicine. 1992, 54:707–719.
14) HokansonJE,DegoodDE,ForrestMS,BrittainTM:Availability
of avoidance behaviors in modulating vascular-stress responses.
Journal of Personality & Social Psychology. 1971, 19:60–68.
15) Degood DE: Cognitive control factors in vascular stress re-
sponses. Psychophysiology. 1975, 12:399–401.
16) Lovallo WR: Activation patterns to aversive stimulation in
man: Passive exposure versus effort to control. Psychophysi-
ology. 1985, 22:283–291.
17) Light KC, Obrist PA: Cardiovascular response to stress: Effects
of opportunity to avoid, shock experience, and performance
feedback. Psychophysiology. 1980, 17:243–252.
18) Kamarck TW, Manuck SB, Jennings JR: Social support re-
duced cardiovascular reactivity to psychological challenge: A
laboratory model. Psychosomatic Medicine. 1990, 52:42–58.
19) Uchino BN, Cacioppo JT, Kiecolt-Glaser JK: The relationship
between social support and physiological processes: A review
with emphasis on underlying mechanisms and implications for
health. Psychological Bulletin. 1996, 119:488–531.
(32) Gerin W, Litt MD, Deich J, Pickering TG: Self-efficacy as a
component of active coping: Effects on cardiovascular reactiv-
ity. Journal of Psychosomatic Research. 1996, 40:485–493.
(33) Fowles DC: Motivational effects on heart rate and electro-
dermal activity: Implications for research on personality and
psychopathology. Journal of Research on Personality. 1983,
17:48–71.
(34) Fowles DC, Fisher AE, Tranel DT: The heart beats to reward:
The effect of monetary incentive on heart rate. Psychophysi-
ology. 1982, 19:506–513.
(35) Kurki T, Smith NT, Head N, Dec-Silver H, Quinn A: Non-
invasive continuous blood pressure measurement from the fin-
ger: Optimal measurement conditions and factors affecting reli-
ability. Journal of Clinical Monitoring. 1987, 3:6–13.
(36) Smith NT, Wesseling KW, De Witt B: Evaluation of two proto-
type devices producing noninvasive pulsatile calibrated blood
pressure from a finger. Journal of Clinical Monitoring. 1985,
1:17–29.
(37) Gerin W, Pieper C, Pickering TG: Measurement reliability of
cardiovascular reactivity change scores: A comparison of inter-
mittent and continuous methods of assessment. Journal of Psy-
chosomatic Research. 1993, 37:493–501.
(
(
(
(
(
(
(
20) House IS, Landis KR, Umberson D: Social relationships and
health. Science. 1988, 241:540–544.
(38) Bandura A: Self-efficacy: Toward a unifying theory of behavior
change. Psychological Review. 1977, 84:191–215.
(
21) Lepore SJ, Allen KAM, Evans GW: Social support lowers car-
diovascular reactivity to an acute stressor. Psychosomatic Med-
icine. 1993, 55:518–524.
(39) Glynn LM, Christenfeld N, Gerin W: Implications of alterna-
tive methods of computing blood pressure means. Blood Pres-
sure Monitor. 1997, 2:175–178.
(
(
22) Glynn LM, Christenfeld N, Gerin W: Gender, social support,
and cardiovascular responses to stress. Psychosomatic Medi-
cine. 1999, 61:234–242.
23) Berkman LF, Syme SL: Social networks, host resistance, and
mortality: A nine year follow-up study of Alameda County resi-
dents. In Steptoe A, Wardle J (eds), Psychosocial Processes and
Health: A Reader. Cambridge, England: Cambridge University
Press, 1994, 43–67.
24) Hafner RJ: Health differences between married men and wom-
en: The contribution of sex-role stereotyping. Australian &
New Zealand Journal of Family Therapy. 1989, 10:13–19.
25) House JS, Strecher VJ, Metzner HL, Robbins CA: Occupa-
tional stress and health among men and women in the Tecum-
(40) Manuck SB, Kasprowicz AL, Monroe SM, Larkin KT, Kaplan
JR: Psychophysiologic reactivity as a dimension of individual
differences. In Schneiderman N, Weiss SM, Kaufmann PG
(eds), Handbook of Research Methods in Cardiovascular Be-
havioral Medicine. New York: Plenum Press, 1989, 365–382.
(41) Lynch JJ, Long JM, Thomas SA, Malinow KL, Katcher AH: The
effectsoftalkingonthebloodpressureofhypertensiveandnormo-
tensive individuals. Psychosomatic Medicine. 1981, 43:25–33.
(42) Gerin W, Pieper C, Levy R, Pickering TG: Social support in so-
cial interaction: A moderator of cardiovascular reactivity. Psy-
chosomatic Medicine. 1992, 54:324–336.
(
(
(43) Baron RM, Kenny DA: The moderator-mediator variable dis-
tinction in social psychological research: Conceptual, strategic,

Volume 24, Number 2, 2002
Audience Status and Reactivity
131
and statistical considerations. Journal of Personality and Social
Psychology. 1986, 51:1173–1182.
(45) Keppel G: Design and Analysis: A Researcher’s Handbook.
Englewood Cliffs, NJ: Prentice Hall, 1973.
(
44) Manuck SB, Kamarck TW, Kasprowicz AS, Waldstein SR: Sta-
bility and patterning of behaviorally evoked cardiovascular re-
activity. In Blascovich JJ, Katkin ES (eds), Cardiovascular Re-
activity to Psychological Stress & Disease. Washington, DC,
American Psychological Association, 1993, 111–134.
(46) Festinger L: A theory of social comparison processes. Human
Relations. 1954, 7:117–140.
(47) Christenfeld N, Gerin W, Linden W, et al.: Social support ef-
fects on cardiovascular reactivity: Is a stranger as effective as a
friend? Psychosomatic Medicine. 1997, 59:388–398.